Data Network Protects Qinghai-Xizang Plateau

Normal monitoring equipment battles to endure winter in the region, as the temperature could drop from -20 ℃ to -30 ℃. Even if the equipment survived, the data stored in the memory cards cannot be transmitted quickly, because nearly 80 percent of the plateau was not covered by a communication network, let alone 4G signals. Researchers had to repeatedly trek into the mountains and manually retrieve the data.

To tackle these challenges, TPDC developed a satellite Internet of Things (IoT) terminal that can adapt to the extreme cold and high-altitude environment. New equipment includes a high-speed unmanned aerial vehicle (UAV) relay system, a UAV-borne runoff monitoring system, and the "Smart Eagle Eye" intelligent monitoring platform. All these build a three-dimensional "space-air-ground" monitoring network across the plateau.

At the space level, TPDC allows for the periodic monitoring of the entire Tibetan Plateau, by integrating domestic and international satellite resources like Gaofen series satellites and Fengyun series satellites.

At the air level, UAVs are equipped with multiple varieties of intelligent payloads to conduct high-resolution and high-frequency monitoring for key areas like glaciers and permafrost, the Yarlung Zangbo Grand Canyon and the Lhalu Wetland.

At the ground level, TPDC deployed its self-developed IoT-based monitoring information system across 24 ground observation stations in six major river basins, continuously collecting in-situ data on 62 key hydrological, meteorological and ecological indicators, with more than two billion data records received to date.

In addition, a dynamic security management mechanism covering the full life cycle of data was established, giving each data set a unique digital object identifier, facilitating the accurate sourcing of data.

Making data "live"

Apart from safely storing the data, TPDC also spared no effort to allow the data to release its scientific values.

It has over 140,000 registered users. In 2025, the data downloaded exceeded 2,100 TB, and the webpage had been visited over 165 million times. In addition, two thirds of the data is fully open and can be downloaded without registration, while 27 percent of the downloads are done internationally.

Feng Min, researcher at ITP and deputy director of TPDC, said TPDC repeatedly ranked first in the assessment of data centers in the geoscience field by third-party evaluation agencies.

TPDC was also the first in the national data center system to introduce a peer review mechanism. Experts open data, examine methods, evaluate precision and give feedback, Feng said, adding that this is like doing academic review for data.

Thanks to the strict quality control, TPDC's data began to play a key role in high-level scientific research.

Taking research on Tibetan Plateau lakes as an example, scientists have a more accurate grasp of the changes in the number, area and volume of lakes with TPDC's data products, generating a series of high-level achievements.

The first set of 75-year long-term near-surface meteorological driver data for China, developed by TPDC, has become a key dataset for research on climate change in the Tibetan Plateau and for engineering safety assessments. The global surface solar radiation data provides a crucial basis for the evaluation of new energy resources.

In addition, the data has been used in national major projects like the construction of the Sichuan-Xizang Railway, hydropower projects in the lower reaches of the Yarlung Zangbo River, and in responses to glacial lake outburst floods.

AI empowers decision making

How to ensure AI truly understands the complicated process of the Earth system remains a challenge, Li said. This is also the direction the TPDC researchers chose to follow.

In terms of data product development, TPDC has established a composite production model integrating multi-source observations, AI-enabled intelligent fusion, and physics-based model assimilation.

The value of data products depends on whether they can be directly used by scientists, Feng said. "AI helps us address efficiency issues in data generation under complex conditions, but every product still requires calibration using real-world data and expert oversight."

TPDC aims at a more challenging goal, modeling and digital twin of the Earth system, and proposed a "unified AI-driven Tibetan Plateau observation, data and model framework."

The model explores the introduction of multi-agent reinforcement learning methods, enabling multiple agents to simulate and project the evolving dynamics of rising temperatures, glacier melting, lake expansion and human adaptation in a virtual environment. This enables decision-making support for Earth system prediction and disaster early warning.

"Different from general large language models, we emphasize more the model to play a role in the prediction of the Earth system and decision making," Li said.

Looking into the future, TPDC is collaborating with other research teams to explore the integration of data assimilation, reinforcement learning and causal analysis within a digital twin of the Earth, with the goal of gradually building a digital twin platform capable of simulating and predicting changes in the plateau Earth system.

"We hope data not only records what happens on the plateau, but also can provide scientific basis for understanding the plateau's changing mechanism and responding to uncertainties in the future," Li said.

Source: Science and Technology Daily